This shows you the differences between two versions of the page.

Link to this comparison view

ntpgps [2018/03/06 08:59]
ntpgps [2020/02/21 12:04] (current)
Line 1: Line 1:
 +====== ntp gps setup ======
 +This process has been complicated and messy, because of the combination of 1) I don't know what I'm doing, and 2) the resources on this are good but complicated. Almost everything I've figured out has come from these three pages:
 +  * http://kenyonralph.com/~kenyon/Time_server/
 +  * http://superuser.com/questions/828036/how-can-i-check-whether-my-ntp-daemon-has-pps-support
 +  * http://www.rjsystems.nl/en/2100-ntpd-garmin-gps-18-lvc-gpsd.php#qrks
 +I had to fill in a few holes here and there, so I am going to try and get a comprehensive guide in place here, because I will forget. 
 +===== Setup =====
 +  * dc7900 running ubu server 14.04
 +  * Garmin 18x LVC to ttyS0 on above, spliced in USB conn for power
 +===== Guide =====
 +  - Install and update ubuntu server on machine. Desktop works but server is lighter. This guide will also probably work for debian but I am writing it for ubu now because that's what I have working.
 +  - [[http://www.techytalk.info/disable-and-remove-apparmor-on-ubuntu-based-linux-distributions/|Disable apparmor]]. This is a single purpose server. Apparmor is garbage and gets in the way of everything. Do so by: <code>sudo service apparmor stop</code><code>sudo update-rc.d -f apparmor remove</code>
 +  - Prepare to build ntp with pps support (get all the bits) <code># apt-get build-dep ntp</code><code> # apt-get install pps-tools ubuntu-dev-tools</code><code># apt-get source ntp</code><code># cd ntp*</code>
 +  - Attempt to build. THIS WILL FAIL! Whatever. If not, who knows. <code># dpkg-buildpackage -uc -us -nc</code>
 +  - It failed. Fix it by editing: <code>nano ntpd/refclock_jupiter.c</code>
 +  - Go to line 904. Comment it out as such <code>/* fprintf(stdout, buffer); */</code> (yes this commenting could be done before the attempted build, but if it works before hand, you save a step. Compiling again after the failure is very fast anyway)
 +  - Compile again <code># dpkg-buildpackage -uc -us -nc</code>
 +  - Install! <code># dpkg -i *.deb</code> (the deb will be up a level)
 +  - Configure udev rules to properly configure the devices at boot. <code># Provide a symlink to /dev/ttyS0 and set low_latency for latency improvement
 +KERNEL=="ttyS0", SYMLINK+="gps0", MODE="0666"
 +KERNEL=="ttyS0", RUN+="/bin/setserial /dev/%k low_latency"
 +KERNEL=="ttyS0", RUN+="/usr/sbin/ldattach pps /dev/%k"
 +# Symlink /dev/pps0 to /dev/gpspps0
 +KERNEL=="pps0", SUBSYSTEM=="pps", DRIVER=="", SYMLINK+="gpspps0"
 +KERNEL=="pps0", GROUP="dialout"
 +KERNEL=="pps0", MODE="0666"</code> For some reason, the ldattach bit never works for me and I have to run it manually after rebooting. This is a huge pain and I don't know how to fix it yet.
 +  - Add user ntp to the dialout group <code>sudo gpasswd -a ntp dialout</code>
 +  - Configure /etc/ntp.conf as such:<code># /etc/ntp.conf, configuration for ntpd; see ntp.conf(5) for help
 +driftfile /var/lib/ntp/ntp.drift
 +# Enable this if you want statistics to be logged.
 +statsdir /var/log/ntpstats/
 +statistics loopstats peerstats clockstats
 +filegen loopstats file loopstats type day enable
 +filegen peerstats file peerstats type day enable
 +filegen clockstats file clockstats type day enable
 +# Specify one or more NTP servers.
 +# local mess
 +server mode 18 minpoll 4 iburst prefer true
 +fudge flag1 1 flag2 0 flag3 1 flag4 1
 +# Use time-a and b ntp servers as a fallback.
 +server time-a.nist.gov
 +server time-b.nist.gov
 +# By default, exchange time with everybody, but don't allow configuration.
 +restrict -4 default kod notrap nomodify nopeer noquery
 +restrict -6 default kod notrap nomodify nopeer noquery
 +# Local users may interrogate the ntp server more closely.
 +restrict ::1
 +# Clients from this subnet have unlimited access, but only if
 +# cryptographically authenticated.
 +restrict mask notrust</code>Tweak as necessary. I don't know all the fudge flags yet. This can be improved.
 +  - Restart ntp service <code>sudo service ntp restart</code>
 +  - Check ntpq -p output. If everything went well, you should see something like this:<code>john@dc7900:~$ ntpq -p
 +     remote           refid      st t when poll reach   delay   offset  jitter
 +oGPS_NMEA(0)     .GPS.            0 l    6   16  377    0.000    0.623   0.072
 +*time-a.nist.gov .ACTS.           1 u   25   64  373   91.647   36.162   4.774
 ++time-b.nist.gov .ACTS.           1 u   18   64  165   96.251   41.469   2.220</code>If anything failed, check journalctl -xe to see if there are any hints. There almost always are.
 +  - Just for funsies, ntpq -pcrv output:<code>john@dc7900:~$ ntpq -pcrv
 +     remote           refid      st t when poll reach   delay   offset  jitter
 +oGPS_NMEA(0)     .GPS.            0 l   15   16  377    0.000    0.492   0.044
 ++time-a.nist.gov .ACTS.           1 u   26   64  337   91.647   36.162   4.632
 +*time-b.nist.gov .ACTS.           1 u   23   64  257   92.135   38.865   4.538
 +associd=0 status=041d leap_none, sync_uhf_radio, 1 event, kern,
 +version="ntpd 4.2.6p5@1.2349-o Fri Feb 26 13:56:57 UTC 2016 (1)",
 +processor="x86_64", system="Linux/4.2.0-16-generic", leap=00, stratum=1,
 +precision=-21, rootdelay=0.000, rootdisp=0.979, refid=GPS,
 +reftime=da7ae733.9e1ec802  Fri, Feb 26 2016 10:01:07.617,
 +clock=da7ae742.600f0f00  Fri, Feb 26 2016 10:01:22.375, peer=45935, tc=4,
 +mintc=3, offset=0.492, frequency=22.723, sys_jitter=0.044,
 +clk_jitter=0.000, clk_wander=0.145</code>
 +====== To D0 & Thoughts ======
 +Right now my offset and jitter can be improved. This can be done by:
 +  * Disable all but GPGGA
 +  * Push pulse width to 200ms
 +  * Some people don't like, and have trouble with, the 18x LVC:
 +    * http://support.ntp.org/bin/view/Support/ConfiguringGarminRefclocks
 +  * Similar to above:
 +    * http://support.ntp.org/bin/view/Support/ConfiguringNMEARefclocks
 +    * I think it's worth noting that my jitter is WAY lower than being reported there. One guy is setting 5-10ms jitter while I am seeing <0.024ms jitter. Either that guy has a bad unit and is being a negative nancy, or firmware updates have made them notably better.
 +  * Get serial pass-through working on the TS140. This will take some tinkering but should be doable. Latency is my only concern.
Driven by DokuWiki Recent changes RSS feed Valid CSS Valid XHTML 1.0